version 3.8-2
- all smoothers available in mgcv now allowed for.
- GEV link function allowed for when using gamlss(). (This is more general, faster and more stable than bgeva.)
- references updated.

version 3.8-1
- manual updated.
- Univariate generalised additive models for location, scale and shape. The new function is gamlss().
- Gaussian distribution with sigma2 corresponding to standard deviation allowed for.
- Other models, links, marginal distributions, copulae included.

version 3.8
- copula additive models dealing with non-random sample selection implemented. The new function is copulaSampleSel().
- Fisk distribution included. Total number of available continuous distributions is now 12.
- trivariate probit models implemented. The new function is SemiParTRIV().
- manual updated.
- various functions polished.

version 3.7
- copula additive models for continuous margins implemented. The new function is copulaReg().
- more link functions available for binary responses.
- manual updated.

version 3.6-1
- starting value procedure for continuous distributions improved.
- At calculation corrected. Thanks to Vanessa Didelez for spotting this.

version 3.6
- Dagum distribution included.
- polys.setup function extended to handle difficult domains.

version 3.5
- AT function extended.
- Three datasets included.
- prev function improved. 
- manual polished.
- OR and RR measures introduced.

version 3.4
- the response for the outcome equation can now be either binary or continuous. 
- starting value procedure for sample selection case simplified.
- starting value procedure for dependence parameter modified.
- post-estimation functions polished a bit.
- partial observability models.

version 3.3
- calculation of bivariate cdfs made faster.
- dependence parameter is now allowed to depend on covariates.
- Student-t case suppressed. Evaluation of cdf was expensive (which did not make this option viable for practical modelling). Moreover, for binary data,
this cdf never really made a difference as compared to a Gaussian cdf.

version 3.2-13.2
- citation file fixed.

version 3.2-13.1
- manual updated.
- Manski's CI included.
- cplot added in summary function.
- Manski's bounds for non-random sample selection case included.

version 3.2-13
- working linear model calculation made more effient.
- alternative, faster and more stable, smoothing parameter procedure implemented.
- worst-case Manski's bound included for comparison.
- gamma measure and odds ratio included in summary function.

version 3.2-12.1
- working linear model calculation made more stable.
- NAs handling fixed.

version 3.2-12
- manual updated and namespace fixed.
- equations have to be provided using a list of two formulas. See examples.
- selection option suppressed and replaced with Model. See documentation. 

version 3.2-11
- post estimation function for sandwich adjustment of the covariance matrix fixed. Thanks to Mark McGovern for reporting the bug.
- asymmetric link functions and penalized fitting implemented.

version 3.2-10
- summary functions improved.
- function to estimate adjusted prevalence from a sample selection model made more efficient.

version 3.2-9
- power probit and reciprocal power probit links.
- function to estimate adjusted prevalence from a sample selection model.

version 3.2-8
- gradient test for the hypothesis of exogeneity (or absence of endogeneity).
- post estimation function for sandwich adjustment of the covariance matrix.
- post estimation function for adjustment of the covariance matrix based on a complex survey design.
- bivariate copula models with nonparametric random effects allowed for.

version 3.2-7
- several bivariate copula models implemented.

version 3.2-6
- examples and references in documentation updated.
- plot and predict functions now based on those available in mgcv.
- starting value procedure for correlation parameter improved.
- main model and post fitting functions written more neatly. 
- AT function improved.

version 3.2-5
- bug in summary function fixed.
- call to smoothing penalty matrices made more efficient.

version 3.2-4
- plot, predict and summary functions rewritten more efficiently.

version 3.2-3
- optional prior weight vector to be used in fitting allowed for.
- prediction function included.
- calculation of joint probabilities in sample selection case updated.

version 3.2-2
- examples in documentation updated.
- command to extract Heckman model estimates fixed.

version 3.2-1
- penalty matrix in lagrange multiplier test fixed.

version 3.2
- lagrange multiplier test (score test) for testing for the presence of endogeneity or sample selection implemented.
- fisher information matrix in sample selection case fixed.
- empirical random effet calculation fixed.
- examples in documentation updated.

version 3.1
- AT function modified to account for estimated random effects.
- automatic smoothing parameter selection for models with nonparametric random effects.
- examples in documentation updated.

version 3.0
- linear predictors extended to include linear, nonlinear and nonparametric random effects.
- p-value computation in summary.SemiParBIVProbit modified.

version 2.0-4.1
- R documentation and references updated.

version 2.0-4
- AT function extended. 
- R documentation and references updated.

version 2.0-3
- two stage procedure for fully parametric fit fixed. 

version 2.0-2
- inflation factor for the UBRE score for univariate and bivariate models. 
- two stage procedure to provide starting values in the sample selection algorithm implemented.

version 2.0
- routine for bivariate probit modelling with sample selection implemented.

version 1.0-2
- function ATE suppressed and replaced by AT.

version 1.0-1
- basis dimension in UBRE smoothing parameter selection function fixed.

version 1.0
- plot function fixed.
- function to calculate residuals from fitted model.

version 0.2
- R documentation and references updated.

version 0.1-1
- initial release.

